| I have been tracking state polls since 1992. In each election since I have found comparing what the state polls are showing against what the national polls are showing to be an enormously useful sanity check. At times the national polls are an early indicator of trends in state polls, since state polls tend to be released a couple of days after they are completed. At other times they point to problems with national polling. Such is the case now.
This table compares the projected national results from state polls against the national polls. I use a four day average to account the occasional day when only a few polls are released. By the way, the numbers of the table below DO NOT include the Big Ten Polls (for reasons described on the flip).
One advantage of this analysis is the sheer volume of state polling. There have been 199 state polls since October 1. You would expect given this volume to see the state polling average be more stable than national polling, and that is fact exactly what we find. Since October 1 Obama's projected lead has ranged between 6 and 8 points. Since October 5th the average has varied by less than 1.2%, though it has trended up in the last few days. This analysis suggests the race has been stable since early October.
In the last two days two polls have suggested a much closer race. More on those two polls on the flip |
| The AP poll found a 1 point lead for Obama, as did today's IBD/Tipp poll. Marc Ambinder identified a significant problem with the AP poll this morning when he noted that 44% of the sample in that poll self identified as born again (much higher than in either 2000 or 2004). The IBD/TIPP poll presents a more interesting case.
To their credit, IBD/TIPP discloses the demographics of their polling. Here is what they disclose:
Democrats: Obama 84, McCain 8
Republicans: McCain 86, Obama 5
Independents: Obama 44, McCain 39
This partisan breakdown isn't really all that different from what other polls are finding. What IS different is that IDP is obviously predicting that more Republicans are going to turnout than Democrats. Rasmussen is currently assuming a partisan breakdown of 39% Democrats, 32% Republicans. If I apply those percentages to the IDP data, I get Obama 47%, McCain 41. Obviously the IDP pollsters do not believe that the Democratic Advantage other pollsters are finding is real.
In point of fact, the Democratic advantage pollsters are finding IS somewhat different from prior elections. The Roper Center's fantastic exit polling archive shows that the partisan breakdown was 37-37 in 2004, 39-35 Democratic in 2000, and 40-35 Democratic in 1996. However, given the polling on questions like the right/track wrong track and the favorability of President Bush, a 7 point Democratic advantage would seem hardly surprising and close to the partisan split in both 2000 and 1996. In fact, when you consider that IBD's own polling shows Obama LEADING among independents, I actually consider the IDB good news for Democrats.
Finally, a word about state polling and pollsters. The table below shows the state polling average and the performance of different pollsters in October.
About 35% of state polling is done by Rasmussen and SurveyUSA, and it is interesting that they are almost identical. Beyond that, it is clear that ARG and Mason-Dixon tend to lean right while PPP and Quinnipiac appear to lean left. Interestingly, Quinnipiac leaned right in 2004.
Finally, a word about the Big Ten Battleground Polling. As I mentioned, I excluded it from this analysis. Big Ten Battleground is a contortion of State Universities that poll their states. One of the pollsters, the University of Wisconsin, has the distinction of being one of the worst pollsters in the business. Their 2004 polling was suspicious and swung between showing Kerry ahead by double digits and showing ahead by low single digits. In addition, I am VERY suspicious of the number in Indiana (which represents a 30 point swing from 2004. It may be that the Big Ten Battleground polling was simply ratifying some of the more optimistic national polling, but until I get confirmation from other pollsters I will exclude it from the averages. |
